Artificial Intelligence, Growth and Employment: The Role of Policy

GAICIL Insights

The AI future is often discussed like the weather, something that happens to societies. Artificial Intelligence, Growth and Employment: The Role of Policy, by Philippe Aghion, Céline Antonin and Simon Bunel, treats it more like architecture.¹ The effects of AI and automation on growth and employment, the authors argue, depend to a large extent on institutions and policies. The paper surveys recent work and presents new empirical evidence from France.

The argument has two parts. First, AI can spur growth by replacing labor, in finite supply, with capital, in unbounded supply, both in producing goods and services and in producing ideas. Yet AI may inhibit growth if combined with inappropriate competition policy. Second, building on Acemoglu and Restrepo, the authors examine robotization's effect on employment in France over 1994-2014. Robotization reduces aggregate employment at the employment zone level, and workers with low education are more negatively affected than educated workers. The promise is unbounded; the pain is local and unevenly shared.

The policy lesson follows. Inappropriate labor market and education policies could reduce the positive impact AI and automation might have on employment. Institutions mediate the technology's effects. Competition policy affects whether AI spurs or inhibits growth; labor market and education policies affect whether workers share the gains or bear the costs. Two societies with identical AI could end up with very different labor markets.

The authors are candid about knowledge gaps. Despite solid theory and compelling empirical work, research is still early in understanding the full welfare implications. Next steps include bridging the analysis of growth and employment and investigating how labor market characteristics affect the nature of innovation, for example whether innovation targets automation or the creation of new product lines.

That framing counters fatalism. Fears about AI often treat its effects as inevitable; this paper insists policy choices shape outcomes. AI's potential to raise growth by relaxing the constraint of finite labor is immense. Realizing it requires competition policy preventing dominant firms from stifling innovation, and education and labor market policies equipping workers to adapt. Destiny, here, is a policy variable.

Competition is the growth hinge. If AI is deployed in markets where incumbents can block entry, growth benefits may be limited. Keeping AI markets open to new innovators lets the technology's potential be realized. Competition policy adapted to AI markets' characteristics is therefore central to growth policy.

Education is the employment hinge. Workers with low education were hit harder by robotization. Investment in education and training helps workers adapt, moving into tasks that complement technology. As AI extends automation to new tasks, these investments matter even more. The robot data from France are a preview of what broader AI automation may do to those left without new skills.

The direction of innovation is the frontier question. Labor market conditions may influence whether firms innovate to automate or to create new products. New product lines can generate new jobs, while automation may displace existing ones. Understanding these choices could help policymakers encourage innovation that expands opportunity. The same research budget can build a replacement or a new market.

Translated into policy, the message empowers. AI's effects depend on decisions they make. Competition policy, education, training and labor market institutions shape whether AI produces broad-based growth and employment or narrower gains. Thoughtful policy can tilt the balance toward inclusive outcomes.

The thread running through it all: technology and institutions interact. AI can spur growth by substituting capital for labor in production and in the generation of ideas, and its employment effects depend on workers' skills and policy design. This paper offers a clear framework and early evidence, making a strong case that the future of AI is a matter of policy choice.

One question to put to anyone forecasting AI's job impact: a forecast that omits competition and education policy may be measuring the policy environment it assumed. Change the institutions, and the same technology may tell a different employment story.